Evaluating cell lines as tumor models by comparison of genomic profiles Domcke, S. et al. Nat. Commun 4:2126.

Evaluating cell lines as tumor models by comparison

of genomic profiles

Domcke, S. et al. Nat. Commun 4:2126

MotivationProblem: Genomic differences between

cancer cell lines and tissue samples

TCGA and CCLE provide molecular profiles for tumor samples and cell lines

Compared high-grade serous ovarian cancer (HGSOC) to genomic profiles to identify suitable cell lines for in vitro models

Ovarian Cancer

Over 100,000 women die of ovarian cancer each year

5th leading cause of cancer deathDivided into 4 major histological subtypes:

Serous (study’s focus)EndometrioidClear CellMucinous carcinoma

Common cell line models for ovarian cancer and HGSOC are: SK-OV-3, A2780, OVCAR-3, CAOV3 and IGROV1

Need for well-characterized cell line models for cell types

Found differences between most common models and majority of HGSOC samples

Analyzed 316 HGSOC tumor samples from TCGA and 47 ovarian cancer cell lines from CCLE

DNA copy-number, mutation and mRNA expression data Fraction genome altered (FGA):

CN=log2(sample intensity/reference intensity)

L(i) is length of segment i

T is threshold value of Cni above which segments are altered

- T= 0.2 for TCGA samples- T=0.3 for CCLE cell lines

Suitability of HGSOC Models

S = A + B – 2×C – D/7

A = Correlation with mean CNA of tumorsB = 1 for cell lines with TP53 mutation or else 0C = 1 for hypermutated cell line or else 0D = # of genes mutated in 7 “non-HGSOC” genes

Future Directions

• Drug response profiles using accurate cell line models with known alterations for patient selection in clinical trials

• Perform preclinical drug screens for more-informed patient therapy

• Any others?